Any recent advances on tresting GOF for closed capture model

[annaren]

I was wondering what the best way to test the GOF for closed capture models. AIC allows me to test the best model out of the candidate models I have selected but this does not necessary mean that it is a good model. I understand the given chi square is not accurate method as thes can be overinflated when the expected values are low.
Can the median c hat be used on closed capture models?

Thanks

[cooch]

I was wondering what the best way to test the GOF for closed capture models. AIC allows me to test the best model out of the candidate models I have selected but this does not necessary mean that it is a good model. I understand the given chi square is not accurate method as thes can be overinflated when the expected values are low.
Can the median c hat be used on closed capture models?

Thanks

Is there something about section 15.6 in the 'closed capture chapter' - which refers explicitly to GOF tests and closed capture models - which isn't clear? I'll assume you *did* actually read this section before posting your question.

[annaren]

I have read this chapter more times than I can remember including section 15.6.
Much earlier someone asked a similar question and was refered to a paper by MacKenzie and Bailey 2004 which uses a form of bootstrapping. I just wondered if there had been any recent advance since then (2004). I know the median c hat is a new approach for open population models and wondered if anythings similar had been devised for closed population models.
Section 15.6 suggests the use of AIC but as I mentioned earlier this only tests the best model out of the candidate models which relies on the best model being in that candidate set.

[annaren]

Sorry but can anybody help me on the previous post (I have read the chapter THOROUGHLY)
Thanks

[cooch]

Sorry but can anybody help me on the previous post (I have read the chapter THOROUGHLY)
Thanks

Generally, if it isn't in the book, or the MARK helpfile, then it doesn't exist within MARK. Simple as that.

As for the larger literature, thats why they invented search engines. And, if it existed in the larger literature, Gary would put it into MARK, and it would then end up in the book (sooner or later). So, there is a non-zero probability that if it isn't in the MARK helpfile, or the book, then it doesn't exist. This is not the case for certain classes of models, but for closed estimators probably reasonable to assume as much.

As noted in the book, the problem is that there is no unambiguous way to define the saturated model (a requirement for generating a robust GOF test). While there are saturated models for some tests, as soon as you get into heterogeneity models, no such luck.